Regulatory search system and method based on word family recommendations

Abstract

The application discloses a kind of based on word family recommendation's regulation and control search system and method, wherein the system includes: acquisition module is used to obtain the text data of preprocessed electric power field;Construction module is used to classify the electric power field vocabulary based on word frequency, and obtain electric power field professional vocabulary;Analysis module is used to construct electric power field word family according to theme description;Mapping module is used to extend electric power field word family according to business label, and obtain electric power field word family with business directionality;Search module is used to search according to supplemented search sentence, and obtain initial search result;Sorting module is used to determine search result according to the arrangement order.It uses pre-training language model and clustering algorithm to carry out electric power field knowledge mining and semantic understanding, realizes the intelligent search of electric power field knowledge and the effective matching of business information, enhances the stability and security of electric power dispatching system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system operation and dispatch, in particular to a dispatch control retrieval system and method based on word family recommendation. BACKGROUND

[0002] In the field of power grid system, with the gradual expansion of the power grid scale, the power network structure is becoming more and more complex, and the huge power network puts forward higher requirements for the power supply quality of the power network. As the core business of power dispatch, the data volume of business data is also increasing, and a large amount of data generated in the process of power dispatch is accumulated. The massive data contains rich value of power dispatch business, and there is a problem of how to develop and utilize these data. Search engine emerges as the times require. Search engine can quickly collect and retrieve information according to certain strategy and algorithm, provide retrieval service for users, and display relevant information to users.

[0003] At present, a large amount of data has been accumulated in the field of power grid dispatch, the data sources are various and the data relations are complex. In the process of preparation, implementation, searching, testing and verification, there is a phenomenon that the standard clauses do not match the business processes. After the annotated text data is disassembled, there is a lack of accurate association with business information, resulting in low searching efficiency of annotated text data. At the same time, the business scenarios involve many systems and different roles, and it is difficult to efficiently associate the fine-grained standard knowledge with system state data and role process information. Business personnel face the problems of low efficiency of manual searching of regulations and operation processes, lack of efficient dispatch information retrieval means, and difficulty in quickly grasping the overall information of power grid operation, which leads to an increase in decision-making time of business personnel in handling accidents.

[0004] In order to avoid the low efficiency and inaccurate query results caused by the mismatch between standard clauses and business processes and the insufficient data association in the process of searching power dispatch business data, the technical personnel in the field have been seeking a method of dispatch control retrieval based on word family recommendation, so as to realize the accurate matching between standard clauses and business processes, improve the association degree and retrieval accuracy of data, and meet the individualized retrieval needs of different role users. SUMMARY

[0005] The purpose of the present application is to provide a dispatch control retrieval system and method based on word family recommendation, which uses pre-trained language models and clustering algorithms for power field knowledge mining and semantic understanding, constructs power field word families, and performs business-oriented expansion to obtain power field knowledge and business association information, realizes intelligent retrieval of power field knowledge and effective matching of business information, improves the level of intelligent information retrieval, reduces the work burden of dispatchers, and enhances the stability and security of the power dispatch system.

[0006] The control retrieval system based on the word family recommendation is characterized in that it comprises:

[0007] The acquisition module is configured to acquire text data in the power field, pre-process the acquired text data in the power field, and obtain pre-processed text data in the power field.

[0008] The construction module is configured to perform word segmentation on the pre-processed text data in the power field by using a word segmentation model, obtain a power field word library, classify the power field word library based on word frequency, and obtain a power field professional word library.

[0009] The analysis module is configured to convert the power field professional word library into a vector by using a pre-trained language model, obtain a word vector of the power field professional word library, perform clustering analysis on the word vector of the power field professional word library by using a clustering algorithm, obtain a word cluster of the power field professional word library, extract a theme description corresponding to the word cluster of the power field professional word library, construct a power field word family according to the theme description, and obtain a word cluster of the power field professional word library.

[0010] The mapping module is configured to formulate a business label according to a business scenario requirement, convert the business label into a vector by using a pre-trained language model, obtain a word vector of the business label, calculate a similarity between the word vector of the business label and the word vector of the power field professional word library, sort the word vector representation of the business label according to the similarity calculation result, filter the word vectors of the business labels after sorting based on the similarity calculation result, obtain a business label corresponding to the word vector of the power field professional word library, expand the power field word family according to the business label, and obtain a power field word family with a business orientation.

[0011] The retrieval module is configured to acquire an input retrieval statement, perform word segmentation on the retrieval statement by using a word segmentation model, find a corresponding word family in the power field word family according to the segmented retrieval statement, supplement the retrieval statement by using the corresponding word family, perform retrieval according to the supplemented retrieval statement, and obtain an initial retrieval result.

[0012] Further, the system described above further comprises a sorting module, which is configured to sort the initial retrieval result according to a preset sorting rule, and determine a retrieval result according to the arrangement order.

[0013] Further, the construction module classifies the power field vocabulary based on word frequency, including: obtaining word frequency data of each word in the power field vocabulary in the power field professional document and in the power field general document, comparing the word frequency data of each word in the power field vocabulary with a preset word frequency threshold, and when the word frequency of the word in the power field professional document is greater than the professional word frequency threshold and the word frequency in the power field general document is less than the general word frequency threshold, the word is added to the power field professional vocabulary.

[0014] Further, the analysis module converts the power field professional vocabulary into vectors through a pre-trained language model, including: defining the power field professional vocabulary; setting the word set in the power field professional vocabulary as W={w1,w2,…,w i};wherein w i represents the i-th word in the word set; using the multi-layer Transformer encoding layer of the pre-trained language model to encode the word w i in the word set W, and the output set of the encoding layer is G={h (1) ,h (2) ,…,h (L)};wherein h (L) represents the output of the L-th encoding layer in the output set, and L represents the number of encoding layers of the pre-trained language model; when calculating the word vector of the word w i , the output h (L) of the last layer of the encoding layer is taken as the contextual semantic representation of the word w i , specifically as follows: Embedding(w i )=h (L) ; the word vector of the word is weighted and fused with the semantic representation of the word to obtain a word vector with context awareness, thereby enhancing the expression ability of the word vector.

[0015] Further, the analysis module performs clustering analysis on the word vectors of the power field professional vocabulary through a clustering algorithm, including: using the word vectors of each word to perform clustering calculation, specifically as follows:

[0016] N ε (v i )={v j ∈R d |||v i -v j ||≤ε}

[0017] wherein N ε represents the neighborhood of the word, v i represents the word vector of the word, v j represents the word vector of the neighbor word, R d represents the vector space, and ||vi -v j || represents the Euclidean distance between word vectors, and epsilon represents the radius of the neighborhood; based on clustering calculation, a plurality of semantically similar word clusters are obtained, and each word cluster contains a plurality of words with similar semantic characteristics in the power field.

[0018] Further, the mapping module calculates the similarity between the word vector of the business label and the word vector of the power field professional word library, which is specifically as follows:

[0019]

[0020] Wherein, similarity(A,B) represents the similarity between the word vector of the business label and the word vector of the power field professional word library, A represents the word vector of the business label; B represents the word vector of the power field professional word library, ‖A‖ represents the length of the word vector of the business label, and ‖B‖ represents the length of the word vector of the power field professional word library.

[0021] Further, the retrieval module finds the corresponding word family in the power field word family according to the segmented retrieval sentence, including: based on the characteristics and logical relationship of the power field word family, hierarchical classification is carried out, and index construction and weight setting are carried out on the power field word family of different levels; the segmented retrieval sentence is matched with the words in the power field word family based on the index and weight, and the synonym, synonym and hypernym in the power field word family are used for expansion matching to obtain the initial retrieval result.

[0022] Further, the sorting module sorts the initial retrieval result according to the preset sorting rule, including: the initial retrieval result is sorted according to the following order: standard item name> standard chapter title> standard chapter text; if the initial retrieval result contains repeated standard item names, it is sorted according to the following order: national standard> enterprise standard> industry standard; if the initial retrieval result only involves standard chapter text, it is sorted in reverse order of word frequency.

[0023] In order to realize the second purpose of the above-mentioned application, the word family recommendation based regulation and control retrieval method is designed, which has the following steps:

[0024] The text data of the power field is obtained, and the obtained text data of the power field is preprocessed to obtain the preprocessed text data of the power field.

[0025] The preprocessed text data of the power field is segmented by a segmentation model to obtain a power field word library; the power field word library is classified based on word frequency to obtain a power field professional word library.

[0026] The power field professional vocabulary is vector-converted by a pre-trained language model to obtain a word vector of the power field professional vocabulary, and the word vector of the power field professional vocabulary is analyzed by a clustering algorithm to obtain a word cluster of the power field professional vocabulary.

[0027] According to the business scenario requirements, a business label is formulated, the business label is vector-converted by a pre-trained language model to obtain a word vector of the business label, the similarity between the word vector of the business label and the word vector of the power field professional vocabulary is calculated, the word vector representation of the business label is sorted according to the similarity calculation result, and the word vector of each business label after sorting is filtered based on the similarity calculation result to obtain a business label corresponding to the word vector of the power field professional vocabulary, the power field word family is expanded according to the business label, and a power field word family with business direction is obtained.

[0028] An input search sentence is obtained, the search sentence is segmented by a segmentation model, the corresponding word family in the power field word family is searched according to the segmented search sentence, the search sentence is supplemented by the corresponding word family, and the initial search result is obtained by searching according to the supplemented search sentence.

[0029] To achieve the above-mentioned purpose of the third aspect of the present application, an electronic device is designed, which comprises:

[0030] At least one processor; and

[0031] The memory is in communication connection with the at least one processor; wherein,

[0032] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method of the present disclosure.

[0033] The present application has the following beneficial effects:

[0034] (1) The regulation and retrieval system and method based on word family recommendation, by pre-training language model for word vector conversion, and using clustering algorithm to analyze the semantic similarity of word vector, the power field professional word cluster is constructed. This semantic-based word family construction method overcomes the problem that synonyms and near-synonyms cannot be effectively matched in traditional keyword retrieval, and improves the coverage and recall rate of retrieval; based on the power field word family, the search sentence input by the user is supplemented, the synonyms, near-synonyms and hypernyms in the word family are used for expansion matching, which avoids the problem that the search sentence is incomplete or inaccurate due to the lack of professional knowledge of the user, and realizes the intelligentization of retrieval.

[0035] (2) The control retrieval system and method based on word family recommendation optimizes the retrieval matching efficiency by classifying the power field word family and indexing and setting the weight of the word family at different levels. At the same time, the expansion of the retrieval sentence by the power field word family also reduces the number of retrieval keywords that the user needs to input; the intelligent retrieval method based on word family recommendation reduces the need for manual intervention or modification of the retrieval sentence. Users do not need to be proficient in professional knowledge to obtain relatively accurate retrieval results, reduce the use threshold, and improve the retrieval efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0036] Fig. 1 A module schematic diagram showing one specific embodiment of a control retrieval system based on word family recommendation of the present application.

[0037] Fig. 2 A flowchart showing one specific embodiment of a control retrieval method based on word family recommendation of the present application. DETAILED DESCRIPTION

[0038] To make the purpose, technical scheme and advantages of the present application clearer, the technical scheme in the present application will be described clearly and completely below in combination with the drawings in the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0039] As Figs. 1-2 shown, the embodiments of the present application disclose a control retrieval system and method based on word family recommendation, which utilizes the characteristics of the text data generated in the historical dispatching retrieval process of the power grid, and obtains scene semantic keywords through text extraction and data analysis, improves the intelligent retrieval level of information, reduces the work burden of dispatchers, and enhances the stability and security of the power dispatching automation system.

[0040] Embodiment 1

[0041] The present embodiment discloses a control retrieval system based on word family recommendation, which comprises:

[0042] The acquisition module is used for acquiring text data in the power field, pre-processing the acquired text data in the power field, and obtaining pre-processed text data in the power field;

[0043] The constructing module is configured to perform word segmentation on the preprocessed text data in the power field by using a word segmentation model, and in specific implementation, Jieba word segmentation can be used to perform word segmentation on the text data; a power field vocabulary is obtained; the power field vocabulary is classified based on word frequency, and a power field professional vocabulary is obtained;

[0044] The analyzing module is configured to perform vector conversion on the power field professional vocabulary by using a pre-trained language model, and in specific implementation, a BERT (Bidirectional Encoder Representations from Transformers) Chinese pre-trained model can be used to vectorize the power field professional vocabulary; a word vector of the power field professional vocabulary is obtained; clustering analysis is performed on the word vector of the power field professional vocabulary by using a clustering algorithm, and in specific implementation, a DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm can be used to perform word vector clustering; a word cluster of the power field professional vocabulary is obtained, and a topic is extracted from the word cluster of the power field professional vocabulary, and in specific implementation, an LDA (Latent Dirichlet Allocation) model can be used to generate a clustering topic; a topic description corresponding to the word cluster of the power field professional vocabulary is obtained, and a power field word family is constructed according to the topic description;

[0045] The mapping module is configured to formulate business labels according to business scenario requirements, for example, “network protection” and “stable calculation”, perform vector conversion on the business labels by using a pre-trained language model, and in specific implementation, a Word2Vec model can be trained by using preprocessed text to vectorize the business labels; a word vector of the business labels is obtained, similarity between the word vector of the business labels and the word vector of the power field professional vocabulary is calculated, the word vector representation of the business labels is sorted by using the similarity calculation result, and each word vector of the business labels after sorting is filtered based on the similarity calculation result, so that a business label corresponding to the word vector of the power field professional vocabulary is obtained, the power field word family is expanded according to the business label, and a power field word family with business orientation is obtained.

[0046] The searching module is configured to obtain an input search statement, perform word segmentation on the search statement by using a word segmentation model, search for a corresponding word family in the power field word family according to the segmented search statement, and in specific implementation, an Elasticsearch distributed search engine can be used for searching; the search statement is supplemented by using the corresponding word family, and searching is performed according to the supplemented search statement, so that an initial search result is obtained.

[0047] Based on the above system, optionally, the system further comprises: a sorting module; the sorting module is used for sorting the initial search results according to a preset sorting rule, and determining the search results according to the arrangement order.

[0048] Specifically, the acquisition module acquires text data in the power field, including terms or labels in power technical dictionaries, power scientific papers, project reports, power regulations or power operation manuals; the acquisition module pre-processes the acquired text data in the power field, including text cleaning, denoising, stop word removal and format unification of the text data.

[0049] Specifically, the construction module classifies the power field vocabulary based on the word frequency, including: acquiring the word frequency data of each word in the power field vocabulary in the power field professional documents and in the power field general documents, comparing the word frequency data of each word in the power field vocabulary with a preset word frequency threshold, and when the word frequency of a word in the power field professional documents is greater than the professional word frequency threshold and the word frequency of the word in the power field general documents is less than the general word frequency threshold, the word is added to the power field professional vocabulary.

[0050] Specifically, the analysis module converts the power field professional vocabulary into vectors through a pre-trained language model, including: defining the power field professional vocabulary; setting the word set in the power field professional vocabulary as W={w1,w2,…,w i};wherein w i represents the i-th word in the word set; using the multi-layer Transformer encoding layer of the pre-trained language model to encode the word w i in the word set W, and the output set of the encoding layer is H={h (1) ,h (2) ,…,h (L)};wherein h (L) represents the output of the L-th encoding layer in the output set, and L represents the number of encoding layers of the pre-trained language model; when calculating the word vector of the word w i , the output h (L) of the last encoding layer is taken as the contextual semantic representation of the word w i , specifically as follows: Embedding(w i )=h (L) ; the word vector of the word is weighted and fused with the semantic representation of the word to obtain a word vector with context awareness, thereby enhancing the expression ability of the word vector.

[0051] Specifically, the analysis module performs clustering analysis on the word vectors of the power field professional vocabulary through a clustering algorithm, including: using the word vectors of each word to perform clustering calculation, specifically as follows:

[0052] N ε (v i )={v j ∈R d |||v i -v j ||≤ε}

[0053] wherein, N ε represents the neighborhood of the word, v i represents the word vector of the word, v j represents the word vector of the neighbor word, R d represents the vector space, ||v i -v j | represents the Euclidean distance between the word vectors, and ε represents the radius of the neighborhood; based on the clustering calculation, a plurality of semantically similar word clusters are obtained, and the result of the clustering is C={C1,C2,…,C m}, wherein m is the number of clusters, C m is a cluster composed of a plurality of words w i , each word cluster contains a plurality of words with similar semantic features in the power field, for example, if the high-frequency words of a certain cluster set are "power grid", "load", and "dispatch", a theme description "power grid load dispatch" can be generated.

[0054] Specifically, the mapping module calculates the similarity between the word vector of the business label and the word vector of the power field professional vocabulary, which is specifically as follows:

[0055]

[0056] wherein, similarity(A,B) represents the similarity between the word vector of the business label and the word vector of the power field professional vocabulary, A represents the word vector of the business label; B represents the word vector of the power field professional vocabulary, ‖A‖ represents the length of the word vector of the business label, and ‖B‖ represents the length of the word vector of the power field professional vocabulary.

[0057] Specifically, the search module searches for a corresponding word family in the power field word family according to the segmented search sentence, including: hierarchical classification based on power field word family characteristics and logical relationships, index construction and weight setting for power field word families at different levels, and optionally including: cold and hot data separation, optimizing data management and resource utilization through different storage strategies; matching the segmented search sentence with the words in the power field word family based on the index and weight, and using the synonyms, near-synonyms and hypernyms in the power field word family for extended matching, for example, the near-synonyms of "electric power" include "electric energy" and "electric power energy", which are similar in meaning to "electric power" and are commonly used to describe related concepts of electric energy; obtaining an initial search result, taking the "network test" word family as an example, the core words are "modeling test", "network protection", "grid-connected performance test", etc., and the related words are "grid source coordination", "mechanical damping", "distributed power supply", etc. When searching for "network test", according to the different levels and weights of various segmented words in the "network test" word family, the similar files containing the words "modeling test", "grid source coordination", etc. are automatically indexed.

[0058] Specifically, the sorting module sorts the initial search results according to a preset sorting rule, including: sorting the initial search results in the following order: standard item name > standard chapter title > standard chapter text; if the initial search results contain duplicate standard item names, sort them in the following order: national standard > enterprise standard > industry standard; if the standard item name contains the words "distribution network", "distribution", "micro-grid", it should be placed after other standard item names; if the initial search results only involve standard chapter text, sort them in descending order of word frequency.

[0059] Embodiment 2

[0060] The embodiment discloses a regulation and control search method based on word family recommendation, which comprises the following steps:

[0061] Step 1, obtaining text data in the power field, preprocessing the obtained text data in the power field to obtain preprocessed text data in the power field;

[0062] Step 2, segmenting the preprocessed text data in the power field through a segmentation model to obtain a power field word library; classifying the power field word library based on word frequency to obtain a power field professional word library;

[0063] Step 3, vector conversion of the power field professional vocabulary library is performed by a pre-trained language model to obtain word vectors of the power field professional vocabulary library; clustering analysis of the word vectors of the power field professional vocabulary library is performed by a clustering algorithm to obtain word clusters of the power field professional vocabulary library, theme extraction is performed on the word clusters of the power field professional vocabulary library to obtain theme descriptions corresponding to the word clusters of the power field professional vocabulary library, and a power field word family is constructed according to the theme descriptions;

[0064] Step 4, business tags are formulated according to business scenario requirements, word vectors of the business tags are obtained by vector conversion of the business tags by a pre-trained language model, similarity between the word vectors of the business tags and the word vectors of the power field professional vocabulary library is calculated, word vector representation of the business tags is sorted using the similarity calculation result, and each business tag whose word vector is sorted is screened based on the similarity calculation result to obtain business tags corresponding to the word vectors of the power field professional vocabulary library, the power field word family is expanded according to the business tags, and a business-oriented power field word family is obtained;

[0065] Step 5, an input search statement is obtained, the search statement is segmented by a segmentation model, corresponding word families are searched in the power field word family according to the segmented search statement, the search statement is supplemented by the corresponding word families, and search is performed according to the supplemented search statement to obtain an initial search result;

[0066] Step 6, the initial search result is sorted according to a preset sorting rule, and a search result is determined according to the sorting order.

[0067] Embodiment 3

[0068] The embodiment discloses an electronic device, and the specific implementation of the electronic device is not limited in the embodiment.

[0069] The electronic device includes a computing unit that can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) or a computer program loaded into a random access memory (RAM) from a storage unit. In the RAM, various programs and data required for the operation of the electronic device can also be stored. The computing unit, the ROM, and the RAM are connected to each other through a bus. An input / output (I / O) interface is also connected to the bus.

[0070] A plurality of components in the electronic device are connected to the I / O interface, including: an input unit, such as a keyboard, a mouse, etc.; an output unit, such as various types of displays, a speaker, etc.; a storage unit, such as a magnetic disk, an optical disk, etc.; and a communication unit, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit allows the electronic device to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunication networks.

[0071] The computing unit can be various general and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit performs various methods and processes described above, such as the data processing method. For example, in some embodiments, the data processing method can be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as the storage unit. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device via the ROM and / or the communication unit. When the computer program is loaded into the RAM and executed by the computing unit, one or more steps of the data processing method described above can be performed. Alternatively, in other embodiments, the computing unit can be configured to perform the data processing method by any other appropriate means, such as by means of firmware.

[0072] Various implementations of the systems and techniques described above herein can be realized in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a programmable logic device (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0073] Program code for carrying out methods of the present disclosure can be written in any combination of one or more programming languages. The program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the program code, when executed by the processor or controller, produces the functions / operations specified in the flowcharts and / or the block diagrams. The program code can be executed entirely on a machine, partially on a machine, partially on a machine and partially on a remote machine or entirely on a remote machine or server. It will be made apparent those of ordinary skill in the art that the specific sequence of steps and / or the arrangement of the described devices and modules can be changed as is devoloped for the convenience and simplicity of explanation without departing from the spirit and scope of the application. The described embodiments and implementations including hardware and software are meant to be examples only and are not limiting as to the scope of the application.

[0074] The algorithms and displays presented herein are not inherently related to any particular computer, virtual system, or other apparatus. Various general purpose systems can be used with programs in accordance with the teachings herein, or it can prove convenient to construct more specialized apparatus to perform the required method steps. The required structure for a variety of these systems will be apparent from the description above. In addition, the present application is not intended to be limited to any particular programming language. It will be appreciated that there are many programming languages that can be used to implement the teachings herein, and any such programming language can be used in connection with the various aspects of the application. The descriptions above are intended to cover all possible implementations of the application, and the application should be interpreted as including all such implementations.

[0075] In the description provided herein, numerous specific details are set forth. However, it is understood that embodiments of the application can be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been described in detail in order to not obscure the understanding of this description.

[0076] Similarly, it is to be understood that the mechanical details of the application can be used in accordance with a variety of embodiments, and that they need not follow necessarily the order described. Also, it is to be understood that the use of certain features can not be intended to imply that such features are essential to the polymeric compositions, methods, and / or articles of manufacture of the present application, unless expressly stated as such.

Claims

1. A regulatory search system based on word family recommendations, characterized by, The method comprises the following steps: An acquisition module is configured to acquire text data in the power field, and preprocess the acquired text data in the power field to obtain preprocessed text data in the power field. A construction module is configured to perform word segmentation on the preprocessed text data in the power field by using a word segmentation model to obtain a power field vocabulary. The power field vocabulary is classified based on word frequency to obtain a power field professional vocabulary. An analysis module is configured to convert the power field professional vocabulary into a vector by using a pre-trained language model to obtain a word vector of the power field professional vocabulary. The word vector of the power field professional vocabulary is analyzed by using a clustering algorithm to obtain a word cluster of the power field professional vocabulary. A mapping module is configured to formulate business tags according to business scenario requirements, convert the business tags into a vector by using a pre-trained language model to obtain a word vector of the business tags, calculate the similarity between the word vector of the business tags and the word vector of the power field professional vocabulary, sort the word vector representation of the business tags based on the similarity calculation result, filter the word vectors of the business tags after sorting based on the similarity calculation result to obtain business tags corresponding to the word vector of the power field professional vocabulary, and expand the power field vocabulary family based on the business tags to obtain a power field vocabulary family with business orientation. A retrieval module is configured to acquire an input search statement, perform word segmentation on the search statement by using a word segmentation model, find a corresponding vocabulary family in the power field vocabulary family based on the segmented search statement, perform hierarchical classification based on the characteristics and logical relationships of the power field vocabulary family, index and set weights for the power field vocabulary family at different levels, match the segmented search statement with the words in the power field vocabulary family based on the index and the weights, and perform expansion matching by using synonyms, near-synonyms and hypernyms in the power field vocabulary family to obtain an initial search result.

2. The system according to claim 1, wherein, Further comprising: An ordering module is configured to order the initial search result according to a preset ordering rule, and determine the search result based on the order.

3. The system of claim 1, wherein, The construction module classifies the power field vocabulary based on word frequency, which comprises the following steps: acquiring the word frequency data of each word in the power field vocabulary in the power field professional documents and in the power field general documents, comparing the word frequency data of each word in the power field vocabulary with a preset word frequency threshold, and adding a word to the power field professional vocabulary when the word frequency of the word in the power field professional documents is greater than a professional word frequency threshold and the word frequency of the word in the power field general documents is less than a general word frequency threshold.

4. The system of claim 1, wherein, The parsing module performs vector transformation on the power industry professional terminology database through a pre-trained language model, including: defining the power industry professional terminology database; assuming the set of words in the power industry professional terminology database is W = {w1, w2, ..., w...} i }; where w i Represent the i-th word in the word set; use the multi-layer Transformer encoding layer of a pre-trained language model to process the words w in the word set W. i Encoding is performed, and the output set of the encoding layer is H = {h} (1) ,h (2) ,…,h (L) }; where h (L) This represents the output of the Lth encoding layer in the output set, where L represents the number of encoding layers in the pre-trained language model; calculating word w i When generating word vectors, the output h of the last encoding layer is used. (L) As a word w i The contextual semantic representation is as follows: Embedding(w i ) = h (L) The word vectors and semantic representations of words are weighted and fused to obtain context-aware word vectors, thereby enhancing the expressive power of word vectors.

5. The system of claim 1, wherein, The analysis module analyzes the word vector of the power field professional vocabulary by using a clustering algorithm, which comprises the following steps: performing clustering calculation by using the word vector of each word, and the calculation is described by the following formula: N ε (v i )={v j ∈R d |‖v i -v j ‖≤ε} where N ε denotes the neighborhood of a word, v i denotes the word vector of a word, v j denotes the word vector of a neighbor word, R d denotes the vector space, ‖v i -v j ‖ denotes the Euclidean distance between word vectors, and ε denotes the radius of the neighborhood; based on clustering calculation, a plurality of word clusters with similar semantics are obtained, and each word cluster contains a plurality of words with similar semantic features in the power field.

6. The system of claim 1, wherein, The mapping module calculates the similarity between the word vector of the business tags and the word vector of the power field professional vocabulary, and the calculation is described by the following formula: Wherein, similarity(A, B) represents the similarity of the word vector of the business label and the word vector of the power field professional vocabulary, A represents the word vector of the business label; B represents the word vector of the power field professional vocabulary, ‖A‖ represents the length of the word vector of the business label, and ‖B‖ represents the length of the word vector of the power field professional vocabulary.

7. The system of claim 2, wherein, The sorting module sorts the initial search results according to a preset sorting rule, including: sorting the initial search results in the following order: standard item name > standard chapter title > standard chapter text; if the initial search results contain repeated standard item names, sorting them in the following order: national standard > enterprise standard > industry standard; if the initial search results only involve standard chapter texts, sorting them in the order of word frequency in reverse order.

8. A method for regulating search based on word family recommendation, characterized in that, The method comprises the following steps: Obtain text data in the power field, and preprocess the obtained text data in the power field to obtain preprocessed text data in the power field; Carry out word segmentation on the preprocessed text data in the power field through a word segmentation model to obtain a power field vocabulary; Classify the power field vocabulary based on word frequency to obtain a power field professional vocabulary; Convert the power field professional vocabulary into a word vector through a pre-trained language model to obtain a word vector of the power field professional vocabulary; Perform clustering analysis on the word vector of the power field professional vocabulary through a clustering algorithm to obtain a word cluster of the power field professional vocabulary, extract a theme description corresponding to the word cluster of the power field professional vocabulary, and construct a power field word family according to the theme description; Formulate a business label according to the business scenario requirements, convert the business label into a word vector through a pre-trained language model, calculate the similarity between the word vector of the business label and the word vector of the power field professional vocabulary, sort the word vector representation of the business label using the similarity calculation result, filter the word vectors of the business labels after sorting based on the similarity calculation result, obtain a business label corresponding to the word vector of the power field professional vocabulary, and expand the power field word family according to the business label to obtain a power field word family with business orientation. Obtain an input search sentence, perform word segmentation on the search sentence through a word segmentation model, find a corresponding word family in the power field word family according to the segmented search sentence, perform hierarchical classification based on the characteristics and logical relationships of the power field word family, index and set weights for the power field word families at different levels, match the segmented search sentence with the words in the power field word family based on the index and weights, and perform expansion matching using the synonyms, near-synonyms and hypernyms in the power field word family to obtain an initial search result.

9. An electronic device, comprising: It comprises: At least one processor; And A memory in communication connection with the at least one processor; wherein The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the word family recommendation-based regulation and control search method in claim 8.

Images